Exploring covariates, rs-fMRI usability, and functional connectivity
Liwei Wang and MB Nebel
2022-03-23
#for flowchart
library(DiagrammeR)
#to export flowchart as .png
#webshot::install_phantomjs()
library(DiagrammeRsvg)
library(rsvg)
library(grid)
library(xtable)
library(dplyr)
library(ggplot2)
library(gridExtra)
#nest/unnest
library(tidyr)
#map function (kind of like a for loop)
library(purrr)
#tidy model summary
library(broom)
library(readxl)
#for tableby
library(arsenal)
#Also uses functions from plyr, scales, mgcv, cowplotDefine geom_split_violin()
Based on https://debruine.github.io/post/plot-comparison/
GeomSplitViolin <- ggproto("GeomSplitViolin", GeomViolin, draw_group = function(self,
data, ..., draw_quantiles = NULL) {
data <- transform(data, xminv = x - violinwidth * (x - xmin), xmaxv = x + violinwidth *
(xmax - x))
grp <- data[1, "group"]
newdata <- plyr::arrange(transform(data, x = if (grp%%2 == 1)
xminv else xmaxv), if (grp%%2 == 1)
y else -y)
newdata <- rbind(newdata[1, ], newdata, newdata[nrow(newdata), ], newdata[1,
])
newdata[c(1, nrow(newdata) - 1, nrow(newdata)), "x"] <- round(newdata[1, "x"])
if (length(draw_quantiles) > 0 & !scales::zero_range(range(data$y))) {
stopifnot(all(draw_quantiles >= 0), all(draw_quantiles <= 1))
quantiles <- ggplot2:::create_quantile_segment_frame(data, draw_quantiles)
aesthetics <- data[rep(1, nrow(quantiles)), setdiff(names(data), c("x", "y")),
drop = FALSE]
aesthetics$alpha <- rep(1, nrow(quantiles))
both <- cbind(quantiles, aesthetics)
quantile_grob <- GeomPath$draw_panel(both, ...)
ggplot2:::ggname("geom_split_violin", grid::grobTree(GeomPolygon$draw_panel(newdata,
...), quantile_grob))
} else {
ggplot2:::ggname("geom_split_violin", GeomPolygon$draw_panel(newdata, ...))
}
})
geom_split_violin <- function(mapping = NULL, data = NULL, stat = "ydensity", position = "identity",
..., draw_quantiles = NULL, trim = TRUE, scale = "area", na.rm = FALSE, show.legend = NA,
inherit.aes = TRUE) {
layer(data = data, mapping = mapping, stat = stat, geom = GeomSplitViolin, position = position,
show.legend = show.legend, inherit.aes = inherit.aes, params = list(trim = trim,
scale = scale, draw_quantiles = draw_quantiles, na.rm = na.rm, ...))
}Load initial data set
load('./Data/noImputation/DataWithPropensities_seed1.RData')
#convert PrimaryDiagnosis to factor
dat3$PrimaryDiagnosis <- factor(dat3$PrimaryDiagnosis, levels = c("Autism", "None"))
levels(dat3$PrimaryDiagnosis)[1] “Autism” “None”
dat3$PrimaryDiagnosis <- relevel(dat3$PrimaryDiagnosis, "None")
levels(dat3$PrimaryDiagnosis) = c("TD", "ASD")
tabInit<- tableby(PrimaryDiagnosis ~ Sex + AgeAtScan,
data=dat3)
summary(tabInit,
title='Summary of diagnosis and sex of all participants who attempted a scan',digits=1, digits.p=4,digits.pct=1, numeric.simplify=TRUE, total=FALSE, test=FALSE)| TD (N=372) | ASD (N=173) | |
|---|---|---|
| Sex | ||
| F | 114 (30.6%) | 25 (14.5%) |
| M | 258 (69.4%) | 148 (85.5%) |
| AgeAtScan | ||
| Mean (SD) | 10.4 (1.2) | 10.4 (1.4) |
| Range | 8.0 - 13.0 | 8.0 - 13.0 |
Our initial cohort is an aggregate of 545 children between 8 and 13-years old who participated in one of several neuroimaging studies at Kennedy Krieger Institute (KKI) between 2007 and 2020.
Reshape data to combine motion quality control (QC) levels
# create dummy factor to include all subjects
dat3$noExclusion <- ifelse(dat3$ID > 0, "Pass", "Pass")
dat3$noExclusion <- factor(dat3$noExclusion, levels = c("Pass", "Fail"))
# convert KKI_criteria to factor with reference level 'Pass'
dat3$KKI_criteria <- factor(dat3$KKI_criteria, levels = c("Pass", "Fail"))
# convert Ciric_length to factor with reference level 'Pass' to match
# KKI_criteria
dat3$Ciric_length <- factor(dat3$Ciric_length, levels = c("Pass", "Fail"))
# combine Ciric_length, KKI, and noExclusion exclusion into one variable
allVariables = c("ID", "PrimaryDiagnosis", "AgeAtScan", "Ciric_length", "KKI_criteria",
"noExclusion", "PANESS.TotalOverflowNotAccountingForAge", "SRS.Score", "WISC.GAI",
"DuPaulHome.InattentionRaw", "DuPaulHome.HyperactivityRaw", "ADOS.Comparable.Total",
"CurrentlyOnStimulants", "HeadCoil", "Sex", "ADHD_Secondary", "SES.Family", "Race2",
"handedness", "CompletePredictorCases", "YearOfScan", "MeanFramewiseDisplacement.KKI")
idVariables = c("ID", "PrimaryDiagnosis", "AgeAtScan", "PANESS.TotalOverflowNotAccountingForAge",
"SRS.Score", "WISC.GAI", "DuPaulHome.InattentionRaw", "DuPaulHome.HyperactivityRaw",
"ADOS.Comparable.Total", "CurrentlyOnStimulants", "HeadCoil", "Sex", "ADHD_Secondary",
"SES.Family", "Race2", "handedness", "CompletePredictorCases", "YearOfScan",
"MeanFramewiseDisplacement.KKI")
qcMelt <- reshape2::melt(dat3[, allVariables], id.vars = names(dat3)[which(names(dat3) %in%
idVariables)], variable.name = "Motion.Exclusion.Level", value.name = "Included")
# rename exclusion levels NOTE: need None to be highest level for
# geom_split_violin
levels(qcMelt$Motion.Exclusion.Level) <- c("Strict", "Lenient", "None")
# convert Included to factor with pass as reference
qcMelt$Included <- factor(qcMelt$Included, levels = c("Pass", "Fail"))
# rename levels of value
levels(qcMelt$Included) <- c("Included", "Excluded")Motion QC levels:
- Strict motion QC = Ciric_length
In the strict case, scans were excluded if mean FD exceeded .2 mm or they included less than five minutes of data free from frames with FD exceeding .25 mm
- Lenient motion QC = KKI_criteria
In the lenient case, scans were excluded if the participant had less than 5 minutes of continuous data after removing frames in which the participant moved more than the nominal size of a voxel between any two frames (3 mm) or their head rotated 3. This procedure was modeled after common head motion exclusion criteria for task fMRI data, which rely on voxel size to determine thresholds for unacceptable motion.
- None = all participants
Limit initial dataset to complete cases
dat3 <- filter(dat3, CompletePredictorCases==1)2.1.4 Determine number of participants in set of complete cases who did not attempt or aborted the scan early
dat3$aborted = rep("No", length=nrow(dat3))
dat3$aborted[is.na(dat3$MeanFramewiseDisplacement) & dat3$KKI_criteria=="Fail"] = "Yes"
tabAbort<- tableby(PrimaryDiagnosis ~ aborted,
data=dat3)
summary(tabAbort,
title='Scans aborted by diagnosis',digits=1, digits.p=4,digits.pct=1, numeric.simplify=TRUE, total=FALSE, test=FALSE)| TD (N=348) | ASD (N=137) | |
|---|---|---|
| aborted | ||
| No | 344 (98.9%) | 134 (97.8%) |
| Yes | 4 (1.1%) | 3 (2.2%) |
Scans were either not attempted after two unsuccessful mock scan training sessions or aborted due to non-compliance for 7 of the participants (3 ASD) in the set of complete cases.
2.1.5 Determine number of participants in complete cases set who attempted more than one scan
load('./Data/nScans.RData')
dat3 <- merge(dat3, nScans, all.x = TRUE)
dat3$n <- factor(dat3$n)
tabNScans<- tableby(PrimaryDiagnosis ~ n,
data=dat3)
summary(tabNScans,
title='Number of scans attempted',digits=1, digits.p=4,digits.pct=1, numeric.simplify=TRUE, total=FALSE, test=FALSE)| TD (N=348) | ASD (N=137) | |
|---|---|---|
| n | ||
| 1 | 282 (81.0%) | 120 (87.6%) |
| 2 | 62 (17.8%) | 17 (12.4%) |
| 3 | 3 (0.9%) | 0 (0.0%) |
| 5 | 1 (0.3%) | 0 (0.0%) |
getmode <- function(v) {
uniqv <- unique(v)
uniqv[which.max(tabulate(match(v, uniqv)))]
}
nMode = getmode(as.numeric(as.character(dat3$n[dat3$n!=1])))83 of the complete cases (66 typically developing, 17 ASD) attempted more than one resting-state fMRI scan. Most participants with multiple attempts had 2 scans.
Table 1. Summarize socio-demographic characteristics of the complete predictor cases for paper
completeCases <- filter(qcMelt, CompletePredictorCases==1)
#make M reference level for sex
completeCases$Sex <- relevel(as.factor(completeCases$Sex), "M")
#labels for table
labels(completeCases) <- c(PrimaryDiagnosis = 'Diagnosis',
AgeAtScan = 'Age in Years',
Sex = 'Sex',
handedness = 'Handedness',
Race2 = 'Race',
SES.Family = 'Socioeconomic Status',
CurrentlyOnStimulants = 'Currently on Stimulants?')
#use chisq for Sex and handedness, kruskal-wallis rank test for Age
tabSex<- tableby( PrimaryDiagnosis~Sex+AgeAtScan+handedness, data=filter(completeCases, Motion.Exclusion.Level=="None" & Included=="Included"), control=tableby.control(numeric.test="kwt", cat.test="chisq", total=FALSE))
#use fisher's exact test for Race, k-w for SES
tabRace<- tableby( PrimaryDiagnosis~Race2+SES.Family, data=filter(completeCases, Motion.Exclusion.Level=="None" & Included=="Included"), control=tableby.control(numeric.test="kwt", cat.test="fe", total=FALSE))
tab12 <- merge(tabSex, tabRace)
#Currently on Stimulants - no test because no TDs are currently on stimulants by design
completeCases$CurrentlyOnStimulants <- factor(completeCases$CurrentlyOnStimulants)
#rename factor levels
levels(completeCases$CurrentlyOnStimulants)[levels(completeCases$CurrentlyOnStimulants)=="0"] <- "No"
levels(completeCases$CurrentlyOnStimulants)[levels(completeCases$CurrentlyOnStimulants)=="1"] <- "Yes"
tabStim<- tableby( PrimaryDiagnosis~CurrentlyOnStimulants, data=filter(completeCases, Motion.Exclusion.Level=="None" & Included=="Included"), control=tableby.control(total=FALSE, test=FALSE))
tab123 <- merge(tab12, tabStim)
summary(tab123)| TD (N=348) | ASD (N=137) | p value | |
|---|---|---|---|
| Sex | 0.002 | ||
| M | 242 (69.5%) | 114 (83.2%) | |
| F | 106 (30.5%) | 23 (16.8%) | |
| Age in Years | 0.664 | ||
| Mean (SD) | 10.353 (1.249) | 10.286 (1.344) | |
| Range | 8.020 - 12.980 | 8.010 - 12.990 | |
| Handedness | 0.308 | ||
| Left | 17 (4.9%) | 10 (7.3%) | |
| Mixed | 19 (5.5%) | 11 (8.0%) | |
| Right | 312 (89.7%) | 116 (84.7%) | |
| Race | 0.005 | ||
| African American | 36 (10.3%) | 7 (5.1%) | |
| Asian | 27 (7.8%) | 3 (2.2%) | |
| Biracial | 45 (12.9%) | 12 (8.8%) | |
| Caucasian | 240 (69.0%) | 115 (83.9%) | |
| Socioeconomic Status | 0.007 | ||
| Mean (SD) | 54.072 (9.408) | 51.883 (9.356) | |
| Range | 18.500 - 66.000 | 27.000 - 66.000 | |
| CurrentlyOnStimulants | |||
| No | 348 (100.0%) | 89 (65.0%) | |
| Yes | 0 (0.0%) | 48 (35.0%) |
Generate version of Table 1 to paste into overleaf
tab <- xtable(as.data.frame(summary(tab123)))
print(tab, type="latex")## % latex table generated in R 4.1.1 by xtable 1.8-4 package
## % Wed Mar 23 13:10:12 2022
## \begin{table}[ht]
## \centering
## \begin{tabular}{rllll}
## \hline
## & & TD (N=348) & ASD (N=137) & p value \\
## \hline
## 1 & **Sex** & & & 0.002 \\
## 2 & \ \ \ M & 242 (69.5\%) & 114 (83.2\%) & \\
## 3 & \ \ \ F & 106 (30.5\%) & 23 (16.8\%) & \\
## 4 & **Age in Years** & & & 0.664 \\
## 5 & \ \ \ Mean (SD) & 10.353 (1.249) & 10.286 (1.344) & \\
## 6 & \ \ \ Range & 8.020 - 12.980 & 8.010 - 12.990 & \\
## 7 & **Handedness** & & & 0.308 \\
## 8 & \ \ \ Left & 17 (4.9\%) & 10 (7.3\%) & \\
## 9 & \ \ \ Mixed & 19 (5.5\%) & 11 (8.0\%) & \\
## 10 & \ \ \ Right & 312 (89.7\%) & 116 (84.7\%) & \\
## 11 & **Race** & & & 0.005 \\
## 12 & \ \ \ African American & 36 (10.3\%) & 7 (5.1\%) & \\
## 13 & \ \ \ Asian & 27 (7.8\%) & 3 (2.2\%) & \\
## 14 & \ \ \ Biracial & 45 (12.9\%) & 12 (8.8\%) & \\
## 15 & \ \ \ Caucasian & 240 (69.0\%) & 115 (83.9\%) & \\
## 16 & **Socioeconomic Status** & & & 0.007 \\
## 17 & \ \ \ Mean (SD) & 54.072 (9.408) & 51.883 (9.356) & \\
## 18 & \ \ \ Range & 18.500 - 66.000 & 27.000 - 66.000 & \\
## 19 & **CurrentlyOnStimulants** & & & \\
## 20 & \ \ \ No & 348 (100.0\%) & 89 (65.0\%) & \\
## 21 & \ \ \ Yes & 0 (0.0\%) & 48 (35.0\%) & \\
## \hline
## \end{tabular}
## \end{table}3.1.1. The impact of motion QC on sample size
Figure 3a. Exclusion flowchart
| Strict (N=485) | Lenient (N=485) | |
|---|---|---|
| Included | ||
| Included | 165 (34.0%) | 390 (80.4%) |
| Excluded | 320 (66.0%) | 95 (19.6%) |
Figure 3a. Motion quality control leads to dramatic reductions in sample size. Flow chart of inclusion criteria for this study showing the number of participants remaining after each exclusion step. Lenient motion quality control (QC) excluded 20% of complete predictor cases, while strict motion QC excluded 66% of complete predictor cases.
Figure 3b. Proportion excluded stratified by Primary Diagnosis and motion QC level
Define theme for proportion excluded plots
My_Theme_prop = theme_light()+theme(
legend.title =element_blank(),
axis.title.x = element_text(size = 12),
axis.title.y = element_text(size = 11),
plot.title = element_text(size = 30),
axis.text.x = element_text(size = 10),
axis.text.y = element_text(size = 10),
strip.text.x = element_text(size = 12,color="black"),
strip.background = element_rect(fill = "white"))Figure 3b. Plot proportions
motion <- filter(completeCases, Motion.Exclusion.Level != "None")
motion$Motion.Exclusion.Level <- droplevels(motion$Motion.Exclusion.Level)
motion <- group_by(motion, PrimaryDiagnosis, Motion.Exclusion.Level, Included)
dx_proportions <- ggplot(motion, aes(x = PrimaryDiagnosis, fill = Included)) + geom_bar(position = "fill",
alpha = 0.6) + facet_grid(~Motion.Exclusion.Level) + scale_fill_manual(values = c("#FDE599",
"#9FB0CC")) + scale_color_manual(values = c("#E9D38D", "#8C9AB4")) + My_Theme_prop +
theme(legend.title = element_blank()) + theme(legend.title = element_blank()) +
ylab("Proportion of Children") + theme(legend.position = "bottom") + theme(legend.margin = margin(t = 0,
r = 0, b = -1, l = -1)) + theme(legend.key.size = unit(0.15, "in"), legend.text = element_text(size = 11)) +
theme(axis.title.x = element_blank()) + theme(axis.text.x = element_text(size = 10))
png("./CovariatesAndRS-fMRIUsability/fig_propExcludedDx_cc.png", width = 3, height = 2.5,
units = "in", res = 200)
dx_proportions
invisible(dev.off())
# Pearson's chi squared tests
extib <- tibble(motion)
exNest <- extib %>%
select(c("PrimaryDiagnosis", "Motion.Exclusion.Level", "Included")) %>%
group_by(Motion.Exclusion.Level) %>%
tidyr::nest()
# nested models
ex_chisq <- exNest %>%
mutate(stats = map(data, ~broom::tidy(chisq.test(.x$PrimaryDiagnosis, .x$Included)))) %>%
unnest(stats)
ex_chisq## # A tibble: 2 × 6
## # Groups: Motion.Exclusion.Level [2]
## Motion.Exclusion.Level data statistic p.value parameter method
## <fct> <list> <dbl> <dbl> <int> <chr>
## 1 Strict <tibble [485 × 2]> 18.3 0.0000186 1 Pears…
## 2 Lenient <tibble [485 × 2]> 8.79 0.00303 1 Pears…dx_proportions
Figure 3b. The proportion of children in each diagnosis group whose scans were included (yellow) and excluded (slate blue) using the strict (left) and lenient (right panel) gross motion QC. A larger proportion of children in the autism spectrum disorder (ASD) group had unusable data and were excluded compared to typically developing (TD) children using lenient motion QC (\(\chi^2\)=8.8, df = 1, p=0.003) and strict (\(\chi^2\)=18.3, df = 1, p=0.003).
Figure 3b. Numbers for main text
tabASD<- tableby(Motion.Exclusion.Level~Included, data=filter(completeCases,
PrimaryDiagnosis=="ASD"))
summary(tabASD,
title = "Proportion of ASD complete cases included/excluded",
digits=0, digits.p=4,digits.pct=1, numeric.simplify=TRUE, total=FALSE, test=FALSE)| Strict (N=137) | Lenient (N=137) | None (N=137) | |
|---|---|---|---|
| Included | |||
| Included | 26 (19.0%) | 98 (71.5%) | 137 (100.0%) |
| Excluded | 111 (81.0%) | 39 (28.5%) | 0 (0.0%) |
tabTD<- tableby(Motion.Exclusion.Level~Included, data=filter(completeCases,
PrimaryDiagnosis=="TD"))
summary(tabTD,
title = "Proportion of TD complete cases included/excluded",
digits=0, digits.p=4,digits.pct=1, numeric.simplify=TRUE, total=FALSE, test=FALSE)| Strict (N=348) | Lenient (N=348) | None (N=348) | |
|---|---|---|---|
| Included | |||
| Included | 139 (39.9%) | 292 (83.9%) | 348 (100.0%) |
| Excluded | 209 (60.1%) | 56 (16.1%) | 0 (0.0%) |
The proportion of children excluded differs across diagnostic groups using both the lenient and strict motion QC.
3.1.2 rs-fMRI exclusion probability changes with phenotype and age
Covariates
phenoVariables <- c("ID", "PrimaryDiagnosis",
"ADOS.Comparable.Total",
"SRS.Score",
"PANESS.TotalOverflowNotAccountingForAge",
"DuPaulHome.InattentionRaw",
"DuPaulHome.HyperactivityRaw",
"AgeAtScan",
"WISC.GAI", "SES.Family",
"Motion.Exclusion.Level", "Included", "Sex")
phenoIDs <- c("ID", "PrimaryDiagnosis", "Motion.Exclusion.Level", "Included", "Sex", "SES.Family")
aim1 <- reshape2::melt(completeCases[, phenoVariables],
id.vars=names(completeCases)[which(names(completeCases) %in% phenoIDs)])## Warning: attributes are not identical across measure variables; they will be
## droppedlevels(aim1$variable) <- c("ADOS", "SRS", "Motor Overflow", "Inattention",
"Hyperactivity", "Age", "GAI")
aim1G <- group_by(aim1, PrimaryDiagnosis, Motion.Exclusion.Level, Included, variable)Fit univarate GAMs.
We used univariate models rather than a model with all covariates simultaneously because some of the variables are correlated, such that the impact of each variable on rs-fMRI usability may be difficult to estimate. These models are related to the propensity models that will be used in the estimation of the deconfounded group difference.
We used univariate models rather than a model with all covariates simultaneously because some of the variables are correlated, such that the impact of each variable on rs-fMRI usability may be difficult to estimate. These models are related to the propensity models that will be used in the estimation of the deconfounded group difference.
aim1$delta = rep(NA,length=nrow(aim1))
aim1$delta = ifelse(aim1$Included=="Included",1,0)
aim1tib <- tibble(filter(aim1, Motion.Exclusion.Level!="None"))
aim1tib$Motion.Exclusion.Level <- droplevels(aim1tib$Motion.Exclusion.Level)
aim1Nest <- aim1tib %>%
group_by(Motion.Exclusion.Level, variable) %>%
tidyr::nest()
#nested models
nested_gams <- aim1Nest %>%
mutate(model = map(data, ~mgcv::gam(1-delta~s(value, k=-1), data = na.omit(.x),
family=binomial(link=logit), method="REML")),
coefs = map(model, tidy, conf.int = FALSE),
Rsq = map_dbl(model, ~summary(.)$r.sq)) %>%
unnest(coefs)
#Ben: correct for 7 lenient and 7 strict
nested_gams_len <- nested_gams %>%
filter(Motion.Exclusion.Level=="Lenient")
nested_gams_len$p.fdr = p.adjust(nested_gams_len$p.value, method = "BH")
nested_gams_strict <- nested_gams %>%
filter(Motion.Exclusion.Level=="Strict")
nested_gams_strict$p.fdr = p.adjust(nested_gams_strict$p.value, method = "BH")
#combine to print
nested_gams <- rbind(nested_gams_len, nested_gams_strict)
#list adjusted p values
nested_gams[, c(1:2,6:11)]## # A tibble: 14 × 8
## # Groups: Motion.Exclusion.Level, variable [14]
## Motion.Exclusion.Level variable edf ref.df statistic p.value Rsq p.fdr
## <fct> <fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 Lenient ADOS 1.54 1.87 15.3 1.44e-3 0.0303 2.77e-3
## 2 Lenient SRS 1.78 2.23 14.9 8.95e-4 0.0411 2.77e-3
## 3 Lenient Motor O… 1.00 1.00 15.6 7.58e-5 0.0308 5.31e-4
## 4 Lenient Inatten… 1.56 1.93 7.83 1.41e-2 0.0167 1.41e-2
## 5 Lenient Hyperac… 1.88 2.36 13.1 2.94e-3 0.0251 4.12e-3
## 6 Lenient Age 1.00 1.00 8.17 4.28e-3 0.0142 4.99e-3
## 7 Lenient GAI 1.00 1.00 9.98 1.59e-3 0.0188 2.77e-3
## 8 Strict ADOS 1.00 1.00 20.3 7.46e-6 0.0426 2.25e-5
## 9 Strict SRS 1.00 1.00 21.0 5.10e-6 0.0547 2.25e-5
## 10 Strict Motor O… 1.00 1.00 10.2 1.42e-3 0.0194 1.99e-3
## 11 Strict Inatten… 1.00 1.00 16.1 5.96e-5 0.0339 1.04e-4
## 12 Strict Hyperac… 1.65 2.05 23.4 9.64e-6 0.0516 2.25e-5
## 13 Strict Age 1.87 2.34 8.05 2.93e-2 0.0139 2.93e-2
## 14 Strict GAI 1.00 1.00 5.90 1.51e-2 0.0101 1.77e-2#max p value for 7 lenient models
max(nested_gams_len$p.fdr)## [1] 0.01409489#max p value for 7 strict models
max(nested_gams_strict$p.fdr)## [1] 0.02926959nested_gams <- nested_gams %>%
mutate(LB = map(data, ~round(min(na.omit(.x$value)))),
UB = map(data, ~round(max(na.omit(.x$value)))),
range = map2(LB, UB, ~seq(from=.x, to=.y, by=1)),
logpredict = map2(model, range, ~predict(.x, newdata = data.frame(value = .y), type="link",se.fit=TRUE)),
fit = map(logpredict, ~plogis(.x$fit)),
lCI = map(logpredict, ~plogis(.x$fit-1.96*.x$se.fit)),
hCI = map(logpredict, ~plogis(.x$fit+1.96*.x$se.fit)))Define theme for Figure 4 top row
gam_theme = theme(
axis.title.x=element_text(size=12),
axis.title.y=element_text(size=12),
axis.text.x=element_text(size=8),
axis.text.y=element_text(size=10),
plot.title = element_text(size = 16),
plot.caption = element_text(size = 16,hjust = 0),
legend.title = element_blank(), legend.position ="none")Figure 4a top. Probability of exclusion as a function of ADOS
ados <- nested_gams %>%
filter(variable=="ADOS") %>%
select("variable", "Motion.Exclusion.Level", "range", "fit", 'lCI', 'hCI') %>%
unnest(c(range, fit, lCI, hCI))
p_ados <- ggplot(ados, aes(x=range, y=fit))+
geom_line(aes(colour = Motion.Exclusion.Level),size=1.2)+ylim(0,1)+theme_bw()+
geom_ribbon(aes(ymin=lCI, ymax=hCI, fill=Motion.Exclusion.Level), linetype='blank', alpha=0.2)+
scale_color_manual(labels=c('Strict', 'Lenient'), values = c("#f55154","#9FB0CC"))+
scale_fill_manual(labels=c('Strict', 'Lenient'), values = c("#f55154","#9FB0CC"))+
labs(x='', y='Probability of Exclusion', fill='Motion Control', colour='Motion Control')+
scale_x_continuous(expand = c(0, 0))+
gam_theme+
ggtitle("ADOS")+
theme(plot.title = element_text(size = 11, hjust = 0.5))Figure 4b top. Probability of exclusion as a function of SRS
srs <- nested_gams %>%
filter(variable=="SRS") %>%
select("variable", "Motion.Exclusion.Level", "range", "fit", 'lCI', 'hCI') %>%
unnest(c(range, fit, lCI, hCI))
p_srs <- ggplot(srs, aes(x=range, y=fit))+
geom_line(aes(colour = Motion.Exclusion.Level),size=1.2)+ylim(0,1)+theme_bw()+
geom_ribbon(aes(ymin=lCI, ymax=hCI, fill=Motion.Exclusion.Level), linetype='blank', alpha=0.2)+
scale_color_manual(labels=c('Strict', 'Lenient'), values = c("#f55154","#9FB0CC"))+
scale_fill_manual(labels=c('Strict', 'Lenient'), values = c("#f55154","#9FB0CC"))+
labs(x='', y='', fill='Motion Control', colour='Motion Control')+
scale_x_continuous(expand = c(0, 0))+
gam_theme+
ggtitle("SRS")+
theme(plot.title = element_text(size = 11, hjust = 0.5))+
theme(axis.title.y = element_blank())Figure 4c top. Probability of exclusion as a function of Inattention
ina <- nested_gams %>%
filter(variable=="Inattention") %>%
select("variable", "Motion.Exclusion.Level", "range", "fit", 'lCI', 'hCI') %>%
unnest(c(range, fit, lCI, hCI))
p_in <- ggplot(ina, aes(x=range, y=fit))+
geom_line(aes(colour = Motion.Exclusion.Level),size=1.2)+ylim(0,1)+theme_bw()+
geom_ribbon(aes(ymin=lCI, ymax=hCI, fill=Motion.Exclusion.Level), linetype='blank', alpha=0.2)+
scale_color_manual(labels=c('Strict', 'Lenient'), values = c("#f55154","#9FB0CC"))+
scale_fill_manual(labels=c('Strict', 'Lenient'), values = c("#f55154","#9FB0CC"))+
labs(x='', y='', fill='Motion Control', colour='Motion Control')+
scale_x_continuous(expand = c(0, 0))+
gam_theme+
ggtitle("Inattention")+
theme(plot.title = element_text(size = 11, hjust = 0.5))+
theme(axis.title.y = element_blank())Figure 4d top. Probability of exclusion as a function of Hyperactivity
hi <- nested_gams %>%
filter(variable=="Hyperactivity") %>%
select("variable", "Motion.Exclusion.Level", "range", "fit", 'lCI', 'hCI') %>%
unnest(c(range, fit, lCI, hCI))
p_hi <- ggplot(hi, aes(x=range, y=fit))+
geom_line(aes(colour = Motion.Exclusion.Level),size=1.2)+ylim(0,1)+theme_bw()+
geom_ribbon(aes(ymin=lCI, ymax=hCI, fill=Motion.Exclusion.Level), linetype='blank', alpha=0.2)+
scale_color_manual(labels=c('Strict', 'Lenient'), values = c("#f55154","#9FB0CC"))+
scale_fill_manual(labels=c('Strict', 'Lenient'), values = c("#f55154","#9FB0CC"))+
labs(x='', y='', fill='Motion Control', colour='Motion Control')+
scale_x_continuous(expand = c(0, 0))+
gam_theme+
ggtitle("Hyperactivity")+
theme(plot.title = element_text(size = 11, hjust = 0.5))+
theme(axis.title.y = element_blank())Figure 4e top. Probability of exclusion as a function of Motor Overflow
mo <- nested_gams %>%
filter(variable=="Motor Overflow") %>%
select("variable", "Motion.Exclusion.Level", "range", "fit", 'lCI', 'hCI') %>%
unnest(c(range, fit, lCI, hCI))
p_mo <- ggplot(mo, aes(x=range, y=fit))+
geom_line(aes(colour = Motion.Exclusion.Level),size=1.2)+ylim(0,1)+theme_bw()+
geom_ribbon(aes(ymin=lCI, ymax=hCI, fill=Motion.Exclusion.Level), linetype='blank', alpha=0.2)+
scale_color_manual(labels=c('Strict', 'Lenient'), values = c("#f55154","#9FB0CC"))+
scale_fill_manual(labels=c('Strict', 'Lenient'), values = c("#f55154","#9FB0CC"))+
labs(x='', y='', fill='Motion Control', colour='Motion Control')+
scale_x_continuous(expand = c(0, 0))+
gam_theme+
ggtitle("Motor Overflow")+
theme(plot.title = element_text(size = 11, hjust = 0.5))+
theme(axis.title.y = element_blank())Figure 4f top. Probability of exclusion as a function of Age
age<- nested_gams %>%
filter(variable=="Age") %>%
select("variable", "Motion.Exclusion.Level", "range", "fit", 'lCI', 'hCI') %>%
unnest(c(range, fit, lCI, hCI))
p_age <- ggplot(age, aes(x=range, y=fit))+
geom_line(aes(colour = Motion.Exclusion.Level),size=1.2)+ylim(0,1)+theme_bw()+
geom_ribbon(aes(ymin=lCI, ymax=hCI, fill=Motion.Exclusion.Level), linetype='blank', alpha=0.2)+
scale_color_manual(labels=c('Strict', 'Lenient'), values = c("#f55154","#9FB0CC"))+
scale_fill_manual(labels=c('Strict', 'Lenient'), values = c("#f55154","#9FB0CC"))+
labs(x='', y='', fill='Motion Control', colour='Motion Control')+
scale_x_continuous(expand = c(0, 0))+
gam_theme+
ggtitle("Age")+
theme(plot.title = element_text(size = 11, hjust = 0.5))+
theme(axis.title.y = element_blank())Figure 4g top. Probability of exclusion as a function of GAI
gai <- nested_gams %>%
filter(variable=="GAI") %>%
select("variable", "Motion.Exclusion.Level", "range", "fit", 'lCI', 'hCI') %>%
unnest(c(range, fit, lCI, hCI))
p_gai <- ggplot(gai, aes(x=range, y=fit))+
geom_line(aes(colour = Motion.Exclusion.Level),size=1.2)+ylim(0,1)+theme_bw()+
geom_ribbon(aes(ymin=lCI, ymax=hCI, fill=Motion.Exclusion.Level), linetype='blank', alpha=0.2)+
scale_color_manual(labels=c('Strict', 'Lenient'), values = c("#f55154","#9FB0CC"))+
scale_fill_manual(labels=c('Strict', 'Lenient'), values = c("#f55154","#9FB0CC"))+
labs(x='', y='', fill='Motion Control', colour='Motion Control')+
scale_x_continuous(expand = c(0, 0))+
gam_theme+
ggtitle("GAI")+
theme(plot.title = element_text(size = 11, hjust = 0.5))+
theme(axis.title.y = element_blank())
p_legend = cowplot::get_legend(p_gai + guides(color = guide_legend(nrow = 1))+
theme(legend.position = "bottom", legend.text = element_text(size = 11),
legend.key.size=unit(.15, "in")))Figure 4 bottom row: Density plots of covariates used to fit GAMs (across included & excluded children)
Define theme for density plots of covariates across included and excluded children
Figure 4a bottom. ADOS density
ddata <- nested_gams %>%
filter(variable=="ADOS") %>%
filter(Motion.Exclusion.Level=="Lenient") %>%
select("variable", "Motion.Exclusion.Level", "data", "LB", "UB") %>%
unnest(c(data, LB, UB)) %>%
filter(PrimaryDiagnosis=="ASD")
ddata$PrimaryDiagnosis <- droplevels(ddata$PrimaryDiagnosis)
d_ados=ggplot(ddata, aes(x=value, fill=PrimaryDiagnosis, color=PrimaryDiagnosis))+
geom_density(alpha=0.5, inherit.aes=TRUE)+
scale_x_continuous(expand = c(0, 0), limits = c(ddata$LB[1], ddata$UB[1]), breaks=seq(0, ddata$UB[1], by=5))+
scale_y_continuous(expand = c(0, 0), limits = c(0, .09), breaks=seq(0, .08, by=.02))+
labs(x='', y='Density')+
scale_fill_manual(values = c("#FDE599"))+
scale_color_manual(values = c("#E9D38D"))+
den_themeFigure 4b bottom. SRS density
ddata <- nested_gams %>%
filter(variable=="SRS") %>%
filter(Motion.Exclusion.Level=="Lenient") %>%
select("variable", "Motion.Exclusion.Level", "data") %>%
unnest(data)
d_srs=ggplot(ddata, aes(x=value, fill=PrimaryDiagnosis, color=PrimaryDiagnosis))+
geom_density(alpha=0.5, inherit.aes=TRUE)+
scale_x_continuous(expand = c(0, 0), limits=c(0,max(srs$range)),breaks = seq(0, 100 , by = 50))+
scale_y_continuous(expand = c(0, 0))+
scale_fill_manual(labels=c('TD','ASD'), values = c("#009E73", "#FDE599"))+
scale_color_manual(labels=c('TD','ASD'), values = c("#05634a", "#E9D38D"))+
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.background = element_blank(),
axis.line = element_line(size=.5), panel.border = element_blank())+
den_theme+
theme(axis.title.y = element_blank())+
labs(x='')Figure 4c bottom. Inattention density
ddata <- nested_gams %>%
filter(variable=="Inattention") %>%
filter(Motion.Exclusion.Level=="Lenient") %>%
select("variable", "Motion.Exclusion.Level", "data") %>%
unnest(data)
d_in=ggplot(ddata, aes(x=value, fill=PrimaryDiagnosis, color=PrimaryDiagnosis))+
geom_density(alpha=0.5, inherit.aes=TRUE)+
scale_x_continuous(expand = c(0, 0))+
scale_y_continuous(expand = c(0, 0))+
scale_fill_manual(labels=c('TD','ASD'), values = c("#009E73", "#FDE599"))+
scale_color_manual(labels=c('TD','ASD'), values = c("#05634a", "#E9D38D"))+
labs(x='')+
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.background = element_blank(),
axis.line = element_line(size=.5), panel.border = element_blank())+
theme(axis.title.y = element_blank())+
den_theme+
theme(axis.title.y = element_blank())+
labs(x='')Figure 4d bottom. Hyperactivity/Impulsivity Density
ddata <- nested_gams %>%
filter(variable=="Hyperactivity") %>%
filter(Motion.Exclusion.Level=="Lenient") %>%
select("variable", "Motion.Exclusion.Level", "data") %>%
unnest(data)
d_hi=ggplot(ddata, aes(x=value, fill=PrimaryDiagnosis, color=PrimaryDiagnosis))+
geom_density(alpha=0.5, inherit.aes=TRUE)+
scale_x_continuous(expand = c(0, 0))+
scale_y_continuous(expand = c(0, 0))+
scale_fill_manual(labels=c('TD','ASD'), values = c("#009E73", "#FDE599"))+
scale_color_manual(labels=c('TD','ASD'), values = c("#05634a", "#E9D38D"))+
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.background = element_blank(),
axis.line = element_line(size=.5), panel.border = element_blank())+
den_theme+
theme(axis.title.y = element_blank())+
labs(x='')Figure 4e bottom. Motor Overflow Density
ddata <- nested_gams %>%
filter(variable=="Motor Overflow") %>%
filter(Motion.Exclusion.Level=="Lenient") %>%
select("variable", "Motion.Exclusion.Level", "data") %>%
unnest(data)
d_mo=ggplot(ddata, aes(x=value, fill=PrimaryDiagnosis, color=PrimaryDiagnosis))+
geom_density(alpha=0.5, inherit.aes=TRUE)+
scale_x_continuous(expand = c(0, 0))+
scale_y_continuous(expand = c(0, 0))+
theme(axis.title.y = element_blank())+
scale_fill_manual(labels=c('TD','ASD'), values = c("#009E73", "#FDE599"))+
scale_color_manual(labels=c('TD','ASD'), values = c("#05634a", "#E9D38D"))+
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.background = element_blank(),
axis.line = element_line(size=.5), panel.border = element_blank())+
den_theme+
theme(axis.title.y = element_blank())+
labs(x='')Figure 4f bottom. Age Density
ddata <- nested_gams %>%
filter(variable=="Age") %>%
filter(Motion.Exclusion.Level=="Lenient") %>%
select("variable", "Motion.Exclusion.Level", "data") %>%
unnest(data)
d_age=ggplot(ddata, aes(x=value, fill=PrimaryDiagnosis, color=PrimaryDiagnosis))+
geom_density(alpha=0.5, inherit.aes=TRUE)+
scale_x_continuous(expand = c(0, 0), limits=c(8,13), breaks = seq(8, 13 , by = 1))+
scale_y_continuous(expand = c(0, 0), limits=c(0,.29), breaks=seq(0, .25, by = .05))+
scale_fill_manual(labels=c('TD','ASD'), values = c("#009E73", "#FDE599"))+
scale_color_manual(labels=c('TD','ASD'), values = c("#05634a", "#E9D38D"))+
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.background = element_blank(),
axis.line = element_line(size=.5), panel.border = element_blank())+
den_theme+
theme(axis.title.y = element_blank())+
labs(x='')Figure 4g bottom. GAI Density
ddata <- nested_gams %>%
filter(variable=="GAI") %>%
filter(Motion.Exclusion.Level=="Lenient") %>%
select("variable", "Motion.Exclusion.Level", "data") %>%
unnest(data)
d_gai=ggplot(ddata, aes(x=value, fill=PrimaryDiagnosis, color=PrimaryDiagnosis))+
geom_density(alpha=0.5, inherit.aes=TRUE)+
scale_x_continuous(expand = c(0, 0))+
scale_y_continuous(expand = c(0, 0), limits = c(0, .035), breaks=seq(0., .03, by=.01))+
scale_fill_manual(labels=c('TD','ASD'), values = c("#009E73", "#FDE599"))+
scale_color_manual(labels=c('TD','ASD'), values = c("#05634a", "#E9D38D"))+
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.background = element_blank(),
axis.line = element_line(size=.5), panel.border = element_blank())+
den_theme+
labs(x='')+
theme(axis.title.y = element_blank())
hist_legend = cowplot::get_legend(d_gai + guides(color = guide_legend(nrow = 1))+theme(legend.position = "bottom", legend.text = element_text(size = 11), legend.key.size=unit(.15, "in")))combine gam plots with densities & print
top_row <- cowplot::plot_grid(p_ados, p_srs, p_in, p_hi, p_mo, p_age, p_gai, ncol=7, rel_widths=c(1.18/7, .97/7, .97/7, .97/7, .97/7, .97/7, .97/7))
bottom_row <- cowplot::plot_grid(d_ados, d_srs, d_in, d_hi, d_mo, d_age, d_gai, ncol=7, rel_widths=c(1.18/7, .97/7, .97/7, .97/7, .97/7, .97/7, .97/7))## Warning: Removed 89 rows containing non-finite values (stat_density).png("./CovariatesAndRS-fMRIUsability/fig_probExclusion_allGAM_TD_ASD_cc.png",width=10,height=6,units="in",res=200)
cowplot::plot_grid(p_legend, top_row, NULL, bottom_row, NULL, hist_legend, nrow=6, rel_heights=c(.1, 1, -.01, .5, -.07, .1))
dev.off()## quartz_off_screen
## 2#png("~/Dropbox/Apps/Overleaf/MotionSelectionBias_rsfMRI/Figures/fig_probExclusion_allGAM_TD_ASD_cc.png",width=10,height=6,units="in",res=200)
cowplot::plot_grid(p_legend, top_row, NULL, bottom_row, NULL, hist_legend, nrow=6, rel_heights=c(.1, 1, -.05, .5, -.07, .1))
#dev.off()NOTE: 19 rows = # of participants with missing Motor Overflow scores (imputed for the DRTMLE of the deconfounded group difference)
Figure 4. rs-fMRI exclusion probability changes with phenotype and age but not SES. Univariate analysis of rs-fMRI exclusion probability as a function of participant characteristics. From left to right: Autism Diagnostic Observation Schedule (ADOS), social responsiveness scale (SRS) scores, inattentive symptoms, hyperactive/impulsive symptoms, total motor overflow, age, general ability index (GAI), and socioeconomic status (SES) using the lenient (light blue lines, all FDR-adjusted p<0.01), and strict (red lines) motion quality control (all FDR-adjusted p<0.03). Variable distributions for each diagnosis group (included and excluded scans) are displayed across the bottom panel (TD=typically developing, green; ASD=autism spectrum disorder, yellow).
3.1.3. Phenotype and age representations differ between included and excluded children
3.1.3. Mann-Whitney U tests to compare included vs excluded participants using lenient motion QC (13 tests)
#run lenient tests first
aim1tib <- tibble(aim1)
aim1MW <- aim1tib %>%
filter(Motion.Exclusion.Level=="Lenient") %>%
group_by(PrimaryDiagnosis, variable) %>%
tidyr::nest()
#hypothesis: included children will have less severe symptoms. NOTE: ADOS only collected in ASD (9 tests)
nested_mw_less <- aim1MW %>%
filter(variable %in% c("SRS", "Inattention", "Hyperactivity",
"Motor Overflow")|(variable=="ADOS"&PrimaryDiagnosis=="ASD")) %>%
mutate(mwm = map(data, ~wilcox.test(value~Included, alternative="less", correct = TRUE, data = na.omit(.x))),
idata = map(data, ~filter(., Included=="Included")),
edata = map(data, ~filter(., Included=="Excluded")),
includedMedian = map(idata, ~median(.x$value, na.rm=TRUE)),
excludedMedian = map(edata, ~median(.x$value, na.rm=TRUE)),
coefs = map(mwm, tidy),
Za = map(coefs, ~qnorm(.x$p.value)),
r = map2(Za, data, ~(abs(.x)/sqrt(length(na.omit(.y$value)))))) %>%
unnest(c(coefs, includedMedian, excludedMedian, r))
#hypothesis: included children will be older and have greater GAI (4 tests)
nested_mw_greater<- aim1MW %>%
filter(variable %in% c("Age", "GAI")) %>%
mutate(mwm = map(data, ~wilcox.test(value~Included, alternative="greater", correct = TRUE, data = na.omit(.x))),
idata = map(data, ~filter(., Included=="Included")),
edata = map(data, ~filter(., Included=="Excluded")),
includedMedian = map(idata, ~median(.x$value, na.rm=TRUE)),
excludedMedian = map(edata, ~median(.x$value, na.rm=TRUE)),
coefs = map(mwm, tidy),
Za = map(coefs, ~qnorm(.x$p.value)),
r = map2(Za, data, ~(abs(.x)/sqrt(length(na.omit(.y$value)))))) %>%
unnest(c(coefs, includedMedian, excludedMedian, r))
complete_mw <- rbind(nested_mw_less, nested_mw_greater)
hist(complete_mw$p.value,
main = "Mann-Whitney U test p values Using Lenient Motion QC")
complete_mw$p.fdr <- p.adjust(complete_mw$p.value, method = "BH")
names(complete_mw)[which(names(complete_mw)=="statistic")]="U"
#complete_mw[order(complete_mw$PrimaryDiagnosis, decreasing=TRUE), c(1:2, 7:9, 13)]
#sort by q-value/p.fdr to ease interpretation of fdr adjusted p values
complete_mw[order(complete_mw$p.fdr, decreasing=FALSE), c(1:2, 7:10, 14:15)]## # A tibble: 13 × 8
## # Groups: PrimaryDiagnosis, variable [13]
## PrimaryDiagnosis variable includedMedian excludedMedian U p.value r
## <fct> <fct> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 ASD Motor O… 17 22 1260 9.48e-4 2.65e-1
## 2 ASD SRS 91.2 100 1439 1.22e-2 1.92e-1
## 3 TD Age 10.3 9.86 9757 1.10e-2 1.23e-1
## 4 ASD ADOS 13 16 1522 3.16e-2 1.59e-1
## 5 TD Hyperac… 1 2 6833 2.31e-2 1.07e-1
## 6 TD GAI 116 112. 9490. 2.84e-2 1.02e-1
## 7 ASD GAI 108 101 2280 3.93e-2 1.50e-1
## 8 ASD Age 10.5 9.76 2263 4.68e-2 1.43e-1
## 9 TD Motor O… 11 13 7086. 5.69e-2 8.48e-2
## 10 TD Inatten… 2 2 7743 2.63e-1 3.40e-2
## 11 ASD Hyperac… 11 12 1832. 3.53e-1 3.22e-2
## 12 TD SRS 16 15 4464. 4.96e-1 5.65e-4
## 13 ASD Inatten… 18 16 1936. 5.48e-1 1.02e-2
## # … with 1 more variable: p.fdr <dbl>#for the paper, sort by PrimaryDiagnosis
xtable(complete_mw[order(complete_mw$PrimaryDiagnosis, decreasing=TRUE), c(1:2, 7:8, 14:15)])## % latex table generated in R 4.1.1 by xtable 1.8-4 package
## % Wed Mar 23 13:10:43 2022
## \begin{table}[ht]
## \centering
## \begin{tabular}{rllrrrr}
## \hline
## & PrimaryDiagnosis & variable & includedMedian & excludedMedian & r & p.fdr \\
## \hline
## 1 & ASD & ADOS & 13.00 & 16.00 & 0.16 & 0.07 \\
## 2 & ASD & SRS & 91.25 & 100.00 & 0.19 & 0.05 \\
## 3 & ASD & Motor Overflow & 17.00 & 22.00 & 0.27 & 0.01 \\
## 4 & ASD & Inattention & 18.00 & 16.00 & 0.01 & 0.55 \\
## 5 & ASD & Hyperactivity & 11.00 & 12.00 & 0.03 & 0.42 \\
## 6 & ASD & Age & 10.54 & 9.76 & 0.14 & 0.08 \\
## 7 & ASD & GAI & 108.00 & 101.00 & 0.15 & 0.07 \\
## 8 & TD & SRS & 16.00 & 15.00 & 0.00 & 0.54 \\
## 9 & TD & Motor Overflow & 11.00 & 13.00 & 0.08 & 0.08 \\
## 10 & TD & Inattention & 2.00 & 2.00 & 0.03 & 0.34 \\
## 11 & TD & Hyperactivity & 1.00 & 2.00 & 0.11 & 0.07 \\
## 12 & TD & Age & 10.32 & 9.87 & 0.12 & 0.05 \\
## 13 & TD & GAI & 116.00 & 112.50 & 0.10 & 0.07 \\
## \hline
## \end{tabular}
## \end{table}#xtable(complete_mw[order(complete_mw$p.fdr, decreasing=FALSE), c(1:2, 7:10, 13)])
#aim1p = complete_mw[, c("PrimaryDiagnosis, variable", "includedMedian", "excludedMedian",
# "U", "p.value", "r", "p.fdr")]
aim1p = complete_mw[, c(1:2, 7:8, 10, 15)]
aim1p$Motion.Exclusion.Level = rep("Lenient", nrow(aim1p))9 of the 13 tests have an FDR-adjusted p value < .2. We expect roughly 1.8 of these to be a false positive.
3.1.3 Supplemental Table S2. Mann-Whitney U tests to compare included vs excluded participants using strict motion QC (13 tests)
#run tests using strict motion QC
aim1MW <- aim1tib %>%
filter(Motion.Exclusion.Level=="Strict") %>%
group_by(variable, PrimaryDiagnosis) %>%
tidyr::nest()
#hypothesis: included children will have less severe symptoms
nested_mw_less <- aim1MW %>%
filter(variable %in% c("SRS", "Inattention", "Hyperactivity",
"Motor Overflow")|(variable=="ADOS"&PrimaryDiagnosis=="ASD")) %>%
mutate(mwm = map(data, ~wilcox.test(value~Included, alternative="less", data = na.omit(.x))),
idata = map(data, ~filter(., Included=="Included")),
edata = map(data, ~filter(., Included=="Excluded")),
includedMedian = map(idata, ~median(.x$value, na.rm=TRUE)),
excludedMedian = map(edata, ~median(.x$value, na.rm=TRUE)),
coefs = map(mwm, tidy),
Za = map(coefs, ~qnorm(.x$p.value)),
r = map2(Za, data, ~(abs(.x)/sqrt(length(na.omit(.y$value)))))) %>%
unnest(c(coefs, includedMedian, excludedMedian, r))
#hypothesis: included children will be older and have greater GAI
nested_mw_greater<- aim1MW %>%
filter(variable %in% c("Age", "GAI")) %>%
mutate(mwm = map(data, ~wilcox.test(value~Included, alternative="greater", data = na.omit(.x))),
idata = map(data, ~filter(., Included=="Included")),
edata = map(data, ~filter(., Included=="Excluded")),
includedMedian = map(idata, ~median(.x$value, na.rm=TRUE)),
excludedMedian = map(edata, ~median(.x$value, na.rm=TRUE)),
coefs = map(mwm, tidy),
Za = map(coefs, ~qnorm(.x$p.value)),
r = map2(Za, data, ~(abs(.x)/sqrt(length(na.omit(.y$value)))))) %>%
unnest(c(coefs, includedMedian, excludedMedian, r))
complete_mw <- rbind(nested_mw_less, nested_mw_greater)
hist(complete_mw$p.value,
main = "Mann-Whitney U test p values Using Strict Motion QC")
complete_mw$p.fdr <- p.adjust(complete_mw$p.value, method = "BH")
names(complete_mw)[which(names(complete_mw)=="statistic")]="U"
#sort by q-value/p.fdr to ease interpretation of fdr adjusted p values
complete_mw[order(complete_mw$p.fdr, decreasing=FALSE), c(1:2, 7:10, 14:15)]## # A tibble: 13 × 8
## # Groups: variable, PrimaryDiagnosis [13]
## PrimaryDiagnosis variable includedMedian excludedMedian U p.value r
## <fct> <fct> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 TD Hyperac… 1 2 11884. 0.00163 0.158
## 2 ASD Age 11.0 10.1 1880 0.00829 0.205
## 3 ASD ADOS 12.5 14 1080. 0.0231 0.170
## 4 ASD SRS 84.2 93.5 1080. 0.0231 0.170
## 5 TD SRS 14.2 17 7047 0.0516 0.101
## 6 ASD Motor O… 15 18 1132. 0.0437 0.146
## 7 TD Age 10.4 10.1 16000. 0.0545 0.0859
## 8 TD GAI 116 115 15797 0.0833 0.0742
## 9 TD Motor O… 11 12 13570. 0.149 0.0557
## 10 TD Inatten… 2 2 13598. 0.154 0.0546
## 11 ASD Inatten… 15.5 18 1306 0.227 0.0641
## 12 ASD Hyperac… 11 12 1297 0.212 0.0683
## 13 ASD GAI 108. 107 1491 0.397 0.0223
## # … with 1 more variable: p.fdr <dbl>#for the paper, sort by PrimaryDiagnosis
xtable(complete_mw[order(complete_mw$PrimaryDiagnosis, decreasing=TRUE), c(1:2, 7:8, 14:15)])## % latex table generated in R 4.1.1 by xtable 1.8-4 package
## % Wed Mar 23 13:10:44 2022
## \begin{table}[ht]
## \centering
## \begin{tabular}{rllrrrr}
## \hline
## & PrimaryDiagnosis & variable & includedMedian & excludedMedian & r & p.fdr \\
## \hline
## 1 & ASD & ADOS & 12.50 & 14.00 & 0.17 & 0.08 \\
## 2 & ASD & SRS & 84.25 & 93.50 & 0.17 & 0.08 \\
## 3 & ASD & Motor Overflow & 15.00 & 18.00 & 0.15 & 0.10 \\
## 4 & ASD & Inattention & 15.50 & 18.00 & 0.06 & 0.25 \\
## 5 & ASD & Hyperactivity & 11.00 & 12.00 & 0.07 & 0.25 \\
## 6 & ASD & Age & 11.04 & 10.13 & 0.20 & 0.05 \\
## 7 & ASD & GAI & 107.50 & 107.00 & 0.02 & 0.40 \\
## 8 & TD & SRS & 14.25 & 17.00 & 0.10 & 0.10 \\
## 9 & TD & Motor Overflow & 11.00 & 12.00 & 0.06 & 0.20 \\
## 10 & TD & Inattention & 2.00 & 2.00 & 0.05 & 0.20 \\
## 11 & TD & Hyperactivity & 1.00 & 2.00 & 0.16 & 0.02 \\
## 12 & TD & Age & 10.38 & 10.13 & 0.09 & 0.10 \\
## 13 & TD & GAI & 116.00 & 115.00 & 0.07 & 0.14 \\
## \hline
## \end{tabular}
## \end{table}#xtable(complete_mw[order(complete_mw$p.fdr, decreasing=FALSE), c(1:2, 7:10, 13)])
temp = complete_mw[, c(1:2,7:8,14:15)]
temp$Motion.Exclusion.Level = rep("Strict", nrow(temp))
aim1p <- rbind(aim1p, temp)
aim1p$p.fdr = round(aim1p$p.fdr, 3)
aim1p$p.signif = rep("", nrow(aim1p))
aim1p$p.signif[aim1p$p.fdr<.2]="^"
aim1p$p.signif[aim1p$p.fdr<.1]="*"
aim1p$p.signif[aim1p$p.fdr<.05]="**"
#think I need this for add_pvalue?
aim1p$Included = rep("Included", nrow(aim1p))Figure 5. Split violin plots
My_Theme = theme_light()+theme(
legend.title = element_blank(),
axis.title.x = element_blank(),
axis.title.y = element_blank(),
axis.text.x = element_text(size = 5),
axis.text.y = element_text(size = 8),
strip.text.x = element_text(size = 12, face = "bold", color="black"),
strip.text.y = element_text(size = 10, color="black"),
strip.background = element_rect(fill="white"),
plot.title = element_text(size = 9, hjust = 0.5))Figure 5a. ADOS (ASD only) split violin
NOTE: Missing ADOS scores for one participant evaluated at CARD (no included in complete predictor cases)
ados <- aim1 %>%
filter(PrimaryDiagnosis=="ASD" & variable=="ADOS") %>%
dplyr::select(-c("PrimaryDiagnosis", "variable"))
stat.test <- aim1p %>% filter(PrimaryDiagnosis=="ASD" & variable=="ADOS")
stat.test$group1 = rep("Included", nrow(stat.test))
stat.test$group2 = rep("null model", nrow(stat.test))
stat.test <- select(stat.test, -p.fdr)
stat.test$y.position = c(27, 27)
paper_ados <- ggplot(ados, aes(Motion.Exclusion.Level, value, fill = Included, color = Included)) +
geom_split_violin(trim=TRUE, alpha = 0.5, inherit.aes = TRUE, adjust = 1.5) +
stat_summary(fun = "mean", position = position_dodge(width = 0.5),
color="black", geom="point", aes(y=value))+
stat_summary(fun.data = "mean_cl_boot", position = position_dodge(width = 0.5),
color="black", geom="errorbar", width=.2)+
#geom_mark_rect(aes(filter = Motion.Exclusion.Level =="Lenient"))+
ggprism::add_pvalue(stat.test,
x = "Motion.Exclusion.Level",
y.position = "y.position",
color="black",
label.size = 6)+
scale_y_continuous(limits=c(0,30),breaks = seq(0, 30 , by = 10))+
scale_fill_manual(values = c("#FDE599", "#9FB0CC"))+
scale_color_manual(values = c("#E9D38D", "#8C9AB4"))+
My_Theme+
theme(legend.text = element_text(size = 8))+
theme(legend.position = "none")+
ggtitle("ADOS")Checking the positivity assumption for DRTMLE (reported in Discussion section of the paper)
adosMaxAll = ados %>% filter(Motion.Exclusion.Level=="Lenient" & Included=="Excluded") %>% select(value) %>% max()
adosMaxUsable = ados %>% filter(Motion.Exclusion.Level=="Lenient" & Included=="Included") %>% select(value) %>% max()The highest ADOS score among included children using lenient motion QC is 23; among all children, 26.
Figure 5b. SRS split violin
srs <- filter(aim1G, variable=="SRS")
stat.test <- aim1p %>% filter(variable=="SRS")
stat.test$group1 = rep("Included", nrow(stat.test))
stat.test$group2 = rep("null model", nrow(stat.test))
stat.test <- select(stat.test, -p.fdr)
stat.test$y.position = c(142, 70, 142, 70)
aim1_srs <- ggplot(srs, aes(Motion.Exclusion.Level, value, fill = Included, color=Included)) +
geom_split_violin(trim=TRUE, alpha = 0.5, inherit.aes = TRUE) +
facet_grid(PrimaryDiagnosis~., scales = "fixed")+
stat_summary(fun = "mean", position = position_dodge(width = 0.5),
color="black", geom="point", aes(y=value))+
stat_summary(fun.data = "mean_cl_boot", position = position_dodge(width = 0.5),
color="black", geom="errorbar", width=.2)+
ggprism::add_pvalue(stat.test,
x = "Motion.Exclusion.Level",
y.position = "y.position",
color="black",
label.size = 6)+
scale_y_continuous(limits=c(0, 150),breaks = seq(0, 100 , by = 50))+
# geom_jitter(position=position_jitterdodge(jitter.height = .25), alpha = .3)+
# geom_pointrange(
# data = summary_pheno,
# aes(Motion.Exclusion.Level, mean, ymin=min, ymax=max),
# shape = 20,
# position = position_dodge(width = 0.9))+
scale_fill_manual(values = c("#FDE599", "#9FB0CC"))+
scale_color_manual(values = c("#E9D38D", "#8C9AB4"))+
My_Theme+
theme(strip.text.y =element_blank())+
theme(legend.position = "none")+
ggtitle("SRS")
v_legend = cowplot::get_legend(aim1_srs + guides(color = guide_legend(nrow = 2))+
theme(legend.position = "left",
legend.text = element_text(size = 10),
legend.key.size=unit(.1, "in")))## Warning: Removed 267 rows containing non-finite values (stat_ydensity).## Warning: Removed 267 rows containing non-finite values (stat_summary).
## Removed 267 rows containing non-finite values (stat_summary).NOTE: 267/3 = 89, # of participants missing SRS
with(dat3, table(PrimaryDiagnosis, is.na(SRS.Score)))##
## PrimaryDiagnosis FALSE TRUE
## TD 259 89
## ASD 137 0Figure 5c. Inattention Split Violin
inat <- filter(aim1G, variable=="Inattention")
stat.test <- aim1p %>% filter(variable=="Inattention")
stat.test$group1 = rep("Included", nrow(stat.test))
stat.test$group2 = rep("null model", nrow(stat.test))
stat.test <- select(stat.test, -p.fdr)
stat.test$y.position = c(20, 20, 20, 20)
paper_inat <- ggplot(inat, aes(Motion.Exclusion.Level, value, fill = Included, color=Included)) +
geom_split_violin(trim=TRUE, alpha = 0.5, inherit.aes = TRUE) +
facet_grid(PrimaryDiagnosis~.)+
stat_summary(fun = "mean", position = position_dodge(width = 0.5),
color="black", geom="point", aes(y=value))+
stat_summary(fun.data = "mean_cl_boot", position = position_dodge(width = 0.5),
color="black", geom="errorbar", width=.2)+
ggprism::add_pvalue(stat.test,
x = "Motion.Exclusion.Level",
y.position = "y.position",
color="black",
label.size = 6)+
# geom_jitter(position=position_jitterdodge(jitter.height = .25), alpha = .3)+
#geom_pointrange(
# data = summary_pheno,
# aes(Motion.Exclusion.Level, mean, ymin=min, ymax=max),
# shape = 20,
# position = position_dodge(width = 0.9))+
scale_fill_manual(values = c("#FDE599", "#9FB0CC"))+
scale_color_manual(values = c("#E9D38D", "#8C9AB4"))+
My_Theme+
theme(strip.text.y = element_blank())+
theme(legend.position = "none")+
ggtitle("Inattention")Figure 5d. Hyperactivity/Impulsivity Spilt Violin
hyp <- filter(aim1G, variable=="Hyperactivity")
stat.test <- aim1p %>% filter(variable=="Hyperactivity")
stat.test$group1 = rep("Included", nrow(stat.test))
stat.test$group2 = rep("null model", nrow(stat.test))
stat.test <- select(stat.test, -p.fdr)
stat.test$y.position = c(16, 16, 16, 16)
paper_hyp <- ggplot(hyp, aes(Motion.Exclusion.Level, value, fill = Included, color=Included)) +
geom_split_violin(trim=TRUE, alpha = 0.5, inherit.aes = TRUE) +
facet_grid(PrimaryDiagnosis~.)+
stat_summary(fun = "mean", position = position_dodge(width = 0.5),
color="black", geom="point", aes(y=value))+
stat_summary(fun.data = "mean_cl_boot", position = position_dodge(width = 0.5),
color="black", geom="errorbar", width=.2)+
ggprism::add_pvalue(stat.test,
x = "Motion.Exclusion.Level",
y.position = "y.position",
color="black",
label.size = 6)+
# geom_jitter(position=position_jitterdodge(jitter.height = .25), alpha = .3)+
#geom_pointrange(
# data = summary_pheno,
# aes(Motion.Exclusion.Level, mean, ymin=min, ymax=max),
# shape = 20,
# position = position_dodge(width = 0.9))+
scale_fill_manual(values = c("#FDE599", "#9FB0CC"))+
scale_color_manual(values = c("#E9D38D", "#8C9AB4"))+
My_Theme+
theme(strip.text.y = element_blank())+
theme(legend.position = "none")+
ggtitle("Hyperactivity")Figure 5e. Motor Overflow Split Violin
overflow <- filter(aim1G, variable=="Motor Overflow")
stat.test <- aim1p %>% filter(variable=="Motor Overflow")
stat.test$group1 = rep("Included", nrow(stat.test))
stat.test$group2 = rep("null model", nrow(stat.test))
stat.test <- select(stat.test, -p.fdr)
stat.test$y.position = c(32, 31, 32, 31)
aim1_of <- ggplot(overflow, aes(Motion.Exclusion.Level, value, fill = Included, color=Included)) +
geom_split_violin(trim=TRUE, alpha = 0.5, inherit.aes = TRUE) +
facet_grid(PrimaryDiagnosis~.)+
stat_summary(fun = "mean", position = position_dodge(width = 0.5),
color="black", geom="point", aes(y=value))+
stat_summary(fun.data = "mean_cl_boot", position = position_dodge(width = 0.5),
color="black", geom="errorbar", width=.2)+
ggprism::add_pvalue(stat.test,
x = "Motion.Exclusion.Level",
y.position = "y.position",
color="black",
label.size = 6)+
scale_y_continuous(limits=c(0,35),breaks = seq(0, 40 , by = 10))+
# geom_jitter(position=position_jitterdodge(jitter.height = .25), alpha = .3)+
#geom_pointrange(
# data = summary_pheno,
# aes(Motion.Exclusion.Level, mean, ymin=min, ymax=max),
# shape = 20,
# position = position_dodge(width = 0.9))+
scale_fill_manual(values = c("#FDE599", "#9FB0CC"))+
scale_color_manual(values = c("#E9D38D", "#8C9AB4"))+
My_Theme+
theme(strip.text.y =element_blank())+
theme(legend.position = "none")+
ggtitle("Motor Overflow")Figure 5f. Age Split Violin
age <- filter(aim1G, variable=="Age")
stat.test <- aim1p %>% filter(variable=="Age")
stat.test$group1 = rep("Included", nrow(stat.test))
stat.test$group2 = rep("null model", nrow(stat.test))
stat.test <- select(stat.test, -p.fdr)
stat.test$y.position = c(13.1, 13.1, 13.1, 13.1)
aim1_age <- ggplot(age, aes(Motion.Exclusion.Level, value, fill = Included, color=Included)) +
geom_split_violin(trim=TRUE, alpha = 0.5, inherit.aes = TRUE) +
facet_grid(PrimaryDiagnosis~.)+
stat_summary(fun = "mean", position = position_dodge(width = 0.5),
color="black", geom="point", aes(y=value))+
stat_summary(fun.data = "mean_cl_boot", position = position_dodge(width = 0.5),
color="black", geom="errorbar", width=.2)+
ggprism::add_pvalue(stat.test,
x = "Motion.Exclusion.Level",
y.position = "y.position",
color="black",
label.size = 6)+
scale_y_continuous(limits=c(8,14),breaks = seq(8, 14 , by = 1))+
# geom_jitter(position=position_jitterdodge(jitter.height = .25), alpha = .3)+
#geom_pointrange(
# data = summary_pheno,
# aes(Motion.Exclusion.Level, mean, ymin=min, ymax=max),
# shape = 20,
# position = position_dodge(width = 0.9))+
scale_fill_manual(values = c("#FDE599", "#9FB0CC"))+
scale_color_manual(values = c("#E9D38D", "#8C9AB4"))+
My_Theme+
theme(strip.text.y =element_blank())+
theme(legend.position = "none")+
ggtitle("Age")Figure 5g. GAI Split Violin
gai <- filter(aim1G, variable=="GAI")
stat.test <- aim1p %>% filter(variable=="GAI")
stat.test$group1 = rep("Included", nrow(stat.test))
stat.test$group2 = rep("null model", nrow(stat.test))
stat.test <- select(stat.test, -p.fdr)
stat.test$y.position = c(155, 160, 155, 160)
aim1_gai <- ggplot(gai, aes(Motion.Exclusion.Level, value, fill = Included, color=Included)) +
geom_split_violin(trim=TRUE, alpha = 0.5, inherit.aes = TRUE) +
facet_grid(PrimaryDiagnosis~.)+
stat_summary(fun = "mean", position = position_dodge(width = 0.5),
color="black", geom="point", aes(y=value))+
stat_summary(fun.data = "mean_cl_boot", position = position_dodge(width = 0.5),
color="black", geom="errorbar", width=.2)+
ggprism::add_pvalue(stat.test,
x = "Motion.Exclusion.Level",
y.position = "y.position",
color="black",
label.size = 6)+
scale_y_continuous(limits=c(75,165),breaks = seq(80, 160, by = 20))+
# geom_jitter(position=position_jitterdodge(jitter.height = .25), alpha = .3)+
#geom_pointrange(
# data = summary_pheno,
# aes(Motion.Exclusion.Level, mean, ymin=min, ymax=max),
# shape = 20,
# position = position_dodge(width = 0.9))+
scale_fill_manual(values = c("#FDE599", "#9FB0CC"))+
scale_color_manual(values = c("#E9D38D", "#8C9AB4"))+
My_Theme+
theme(legend.position = "none")+
ggtitle("GAI")Combine split violins into one figure & save
Figure 5. Participants with usable rs-fMRI data differed from participants with unusable rs-fMRI data. Comparison of Autism Diagnostic Observation Schedule (ADOS) scores, social responsiveness scale (SRS) scores, inattentive symptoms, hyperactive/impulsive symptoms, motor overflow, age, and general ability index (GAI) for included (yellow) and excluded (slate blue) participants stratified by diagnosis group and motion exclusion level. The deconfounded mean integrates across the diagnosis-specific distribution of usable and unusable covariates, which here is labeled as None. Mean values are indicated by a black dot; 95% bootstrap confidence intervals are indicated with black bars. We controlled for 13 comparisons performed for the lenient and strict motion QC cases using the false discovery rate (FDR). ** indicate differences between included and excluded participants with an FDR-adjusted p value <0.05; * indicate FDR-adjusted p values <0.1. ^ indicate FDR-adjusted p values <0.2. A larger number of significant differences are observed using the lenient motion QC than the strict motion QC, but very few participants pass strict motion QC. autism spectrum disorder (ASD), typically developing (TD).
3.1.4. Functional connectivity as a function of phenotype and age
Combine partial correlations with melted rs-fMRI usability and covariate info
startEdgeidx = which(names(dat3)=='r.ic1.ic2')
endEdgeidx = which(names(dat3)=='r.ic29.ic30')
names(dat3)[startEdgeidx:endEdgeidx]## [1] "r.ic1.ic2" "r.ic1.ic4" "r.ic1.ic8" "r.ic1.ic13" "r.ic1.ic14"
## [6] "r.ic1.ic15" "r.ic1.ic17" "r.ic1.ic19" "r.ic1.ic21" "r.ic1.ic22"
## [11] "r.ic1.ic24" "r.ic1.ic25" "r.ic1.ic26" "r.ic1.ic27" "r.ic1.ic28"
## [16] "r.ic1.ic29" "r.ic1.ic30" "r.ic2.ic4" "r.ic2.ic8" "r.ic2.ic13"
## [21] "r.ic2.ic14" "r.ic2.ic15" "r.ic2.ic17" "r.ic2.ic19" "r.ic2.ic21"
## [26] "r.ic2.ic22" "r.ic2.ic24" "r.ic2.ic25" "r.ic2.ic26" "r.ic2.ic27"
## [31] "r.ic2.ic28" "r.ic2.ic29" "r.ic2.ic30" "r.ic4.ic8" "r.ic4.ic13"
## [36] "r.ic4.ic14" "r.ic4.ic15" "r.ic4.ic17" "r.ic4.ic19" "r.ic4.ic21"
## [41] "r.ic4.ic22" "r.ic4.ic24" "r.ic4.ic25" "r.ic4.ic26" "r.ic4.ic27"
## [46] "r.ic4.ic28" "r.ic4.ic29" "r.ic4.ic30" "r.ic8.ic13" "r.ic8.ic14"
## [51] "r.ic8.ic15" "r.ic8.ic17" "r.ic8.ic19" "r.ic8.ic21" "r.ic8.ic22"
## [56] "r.ic8.ic24" "r.ic8.ic25" "r.ic8.ic26" "r.ic8.ic27" "r.ic8.ic28"
## [61] "r.ic8.ic29" "r.ic8.ic30" "r.ic13.ic14" "r.ic13.ic15" "r.ic13.ic17"
## [66] "r.ic13.ic19" "r.ic13.ic21" "r.ic13.ic22" "r.ic13.ic24" "r.ic13.ic25"
## [71] "r.ic13.ic26" "r.ic13.ic27" "r.ic13.ic28" "r.ic13.ic29" "r.ic13.ic30"
## [76] "r.ic14.ic15" "r.ic14.ic17" "r.ic14.ic19" "r.ic14.ic21" "r.ic14.ic22"
## [81] "r.ic14.ic24" "r.ic14.ic25" "r.ic14.ic26" "r.ic14.ic27" "r.ic14.ic28"
## [86] "r.ic14.ic29" "r.ic14.ic30" "r.ic15.ic17" "r.ic15.ic19" "r.ic15.ic21"
## [91] "r.ic15.ic22" "r.ic15.ic24" "r.ic15.ic25" "r.ic15.ic26" "r.ic15.ic27"
## [96] "r.ic15.ic28" "r.ic15.ic29" "r.ic15.ic30" "r.ic17.ic19" "r.ic17.ic21"
## [101] "r.ic17.ic22" "r.ic17.ic24" "r.ic17.ic25" "r.ic17.ic26" "r.ic17.ic27"
## [106] "r.ic17.ic28" "r.ic17.ic29" "r.ic17.ic30" "r.ic19.ic21" "r.ic19.ic22"
## [111] "r.ic19.ic24" "r.ic19.ic25" "r.ic19.ic26" "r.ic19.ic27" "r.ic19.ic28"
## [116] "r.ic19.ic29" "r.ic19.ic30" "r.ic21.ic22" "r.ic21.ic24" "r.ic21.ic25"
## [121] "r.ic21.ic26" "r.ic21.ic27" "r.ic21.ic28" "r.ic21.ic29" "r.ic21.ic30"
## [126] "r.ic22.ic24" "r.ic22.ic25" "r.ic22.ic26" "r.ic22.ic27" "r.ic22.ic28"
## [131] "r.ic22.ic29" "r.ic22.ic30" "r.ic24.ic25" "r.ic24.ic26" "r.ic24.ic27"
## [136] "r.ic24.ic28" "r.ic24.ic29" "r.ic24.ic30" "r.ic25.ic26" "r.ic25.ic27"
## [141] "r.ic25.ic28" "r.ic25.ic29" "r.ic25.ic30" "r.ic26.ic27" "r.ic26.ic28"
## [146] "r.ic26.ic29" "r.ic26.ic30" "r.ic27.ic28" "r.ic27.ic29" "r.ic27.ic30"
## [151] "r.ic28.ic29" "r.ic28.ic30" "r.ic29.ic30"signalFC <- dat3[, c(1,startEdgeidx:endEdgeidx)]
dat2 <- merge(aim1, signalFC, all=TRUE, by = "ID")
fcMelt <- reshape2::melt(dat2[,1:162],
id.vars=names(dat2)[1:9],
variable.name = "edge",
value.name = "fc")3.1.4. Run nested gams
fctib <- tibble(filter(fcMelt, Motion.Exclusion.Level!="None"))
fctib$Motion.Exclusion.Level <- droplevels(fctib$Motion.Exclusion.Level)
fcNest <- fctib %>%
filter(Included=="Included") %>%
group_by(variable, Motion.Exclusion.Level, edge) %>%
tidyr::nest()
#nested models
fc_gams <- fcNest %>%
mutate(model = map(data, ~mgcv::gam(fc~s(value), data = na.omit(.x), method="REML")),
coefs = map(model, tidy, conf.int = FALSE)) %>%
unnest(coefs)Figure 6 Plot histograms of edgewise p-values from GAMs
Figure 6a. Histogram of edgewise p-values for partial correlations as a function of ADOS.
NOTE: TD scores = 0
pdat <- fc_gams %>%
ungroup() %>%
filter(variable=="ADOS") %>%
select(Motion.Exclusion.Level, p.value) %>%
group_by(Motion.Exclusion.Level)
hist_ados_p=ggplot(pdat, aes(x=p.value, fill=Motion.Exclusion.Level, color=Motion.Exclusion.Level))+
geom_histogram(position = "identity", alpha=0.5, inherit.aes=TRUE, binwidth = .05)+
scale_x_continuous(expand = c(0, 0))+
scale_y_continuous(expand = c(0, 0))+
scale_color_manual(labels=c('Strict', 'Lenient'), values = c("#f55154", "#9FB0CC"))+
scale_fill_manual(labels=c('Strict', 'Lenient'), values = c("#f55154", "#9FB0CC"))+
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.background = element_blank(),
axis.line = element_line(size=.5), panel.border = element_blank())+
My_Theme+
ggtitle("ADOS")+
theme(plot.title = element_text(size = 9, hjust = 0.5))+
theme(legend.position = "none")+
labs(x='p value', y='Count')+
theme(axis.title.y = element_text(size = 9, angle=90))+
theme(axis.title.x = element_text(size = 7))Figure 6b. Histogram of edgewise p-values for partial correlations as a function of SRS
pdat <- fc_gams %>%
ungroup() %>%
filter(variable=="SRS") %>%
select(Motion.Exclusion.Level, p.value) %>%
group_by(Motion.Exclusion.Level)
hist_srs_p=ggplot(pdat, aes(x=p.value, fill=Motion.Exclusion.Level, color=Motion.Exclusion.Level))+
geom_histogram(position = "identity", alpha=0.5, inherit.aes=TRUE, binwidth = .05)+
scale_x_continuous(expand = c(0, 0))+
scale_y_continuous(expand = c(0, 0))+
labs(x='p value', y='')+
scale_color_manual(labels=c('Strict', 'Lenient'), values = c("#f55154", "#9FB0CC"))+
scale_fill_manual(labels=c('Strict', 'Lenient'), values = c("#f55154", "#9FB0CC"))+
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.background = element_blank(),
axis.line = element_line(size=.5), panel.border = element_blank())+
My_Theme+
theme(axis.title.y = element_blank())+
ggtitle("SRS")+
theme(plot.title = element_text(size = 9, hjust = 0.5))+
theme(legend.position = "none")+
theme(axis.title.x = element_text(size = 7))Figure 6c. Histogram of p-values for partial correlations as a function of inattention
pdat <- fc_gams %>%
ungroup() %>%
filter(variable=="Inattention") %>%
select(Motion.Exclusion.Level, p.value) %>%
group_by(Motion.Exclusion.Level)
hist_in_p=ggplot(pdat, aes(x=p.value, fill=Motion.Exclusion.Level, color=Motion.Exclusion.Level))+
geom_histogram(position = "identity", alpha=0.5, inherit.aes=TRUE, binwidth = .05)+
scale_x_continuous(expand = c(0, 0))+
scale_y_continuous(expand = c(0, 0))+
labs(x='p value', y='')+
scale_color_manual(labels=c('Strict', 'Lenient'), values = c("#f55154", "#9FB0CC"))+
scale_fill_manual(labels=c('Strict', 'Lenient'), values = c("#f55154", "#9FB0CC"))+
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.background = element_blank(),
axis.line = element_line(size=.5), panel.border = element_blank())+
My_Theme+
theme(axis.title.y = element_blank())+
ggtitle("Inattention")+
theme(plot.title = element_text(size = 9, hjust = 0.5))+
theme(legend.position = "none")+
theme(axis.title.x = element_text(size = 7))Number of edges with a p value < 0.05 for SRS: #### Figure 6d. Histogram of p-values for partial correlations as a function of hyperactivity
pdat <- fc_gams %>%
ungroup() %>%
filter(variable=="Hyperactivity") %>%
select(Motion.Exclusion.Level, p.value) %>%
group_by(Motion.Exclusion.Level)
hist_hi_p=ggplot(pdat, aes(x=p.value, fill=Motion.Exclusion.Level, color=Motion.Exclusion.Level))+
geom_histogram(position = "identity", alpha=0.5, inherit.aes=TRUE, binwidth = .05)+
scale_x_continuous(expand = c(0, 0))+
scale_y_continuous(expand = c(0, 0))+
labs(x='p value', y='')+
scale_color_manual(labels=c('Strict', 'Lenient'), values = c("#f55154", "#9FB0CC"))+
scale_fill_manual(labels=c('Strict', 'Lenient'), values = c("#f55154", "#9FB0CC"))+
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.background = element_blank(),
axis.line = element_line(size=.5), panel.border = element_blank())+
My_Theme+
theme(axis.title.y = element_blank())+
ggtitle("Hyperactivity")+
theme(plot.title = element_text(size = 9, hjust = 0.5))+
theme(legend.position = "none")+
theme(axis.title.x = element_text(size = 7))Figure 6e. Histogram of p-values for partial correlations as a function of motor overflow
pdat <- fc_gams %>%
ungroup() %>%
filter(variable=="Motor Overflow") %>%
select(Motion.Exclusion.Level, p.value) %>%
group_by(Motion.Exclusion.Level)
hist_mo_p=ggplot(pdat, aes(x=p.value, fill=Motion.Exclusion.Level, color=Motion.Exclusion.Level))+
geom_histogram(position = "identity", alpha=0.5, inherit.aes=TRUE, binwidth = .05)+
scale_x_continuous(expand = c(0, 0))+
scale_y_continuous(expand = c(0, 0))+
labs(x='p value', y='')+
scale_color_manual(labels=c('Strict', 'Lenient'), values = c("#f55154", "#9FB0CC"))+
scale_fill_manual(labels=c('Strict', 'Lenient'), values = c("#f55154", "#9FB0CC"))+
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.background = element_blank(),
axis.line = element_line(size=.5), panel.border = element_blank())+
My_Theme+
theme(axis.title.y = element_blank())+
ggtitle("Motor Overflow")+
theme(plot.title = element_text(size = 9, hjust = 0.5))+
theme(legend.position = "none")+
theme(axis.title.x = element_text(size = 7))Figure 6f. Histogram of p-values for partial correlations as a function of age
pdat <- fc_gams %>%
ungroup() %>%
filter(variable=="Age") %>%
select(Motion.Exclusion.Level, p.value) %>%
group_by(Motion.Exclusion.Level)
hist_age_p=ggplot(pdat, aes(x=p.value, fill=Motion.Exclusion.Level, color=Motion.Exclusion.Level))+
geom_histogram(position = "identity", alpha=0.5, inherit.aes=TRUE, binwidth = .05)+
scale_x_continuous(expand = c(0, 0))+
scale_y_continuous(expand = c(0, 0))+
labs(x='p value', y='')+
scale_color_manual(labels=c('Strict', 'Lenient'), values = c("#f55154", "#9FB0CC"))+
scale_fill_manual(labels=c('Strict', 'Lenient'), values = c("#f55154", "#9FB0CC"))+
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.background = element_blank(),
axis.line = element_line(size=.5), panel.border = element_blank())+
My_Theme+
theme(axis.title.y = element_blank())+
ggtitle("Age")+
theme(plot.title = element_text(size = 9, hjust = 0.5))+
theme(legend.position = "none")+
theme(axis.title.x = element_text(size = 7))Figure 6g. Histogram of p-values for partial correlations as a function of GAI
pdat <- fc_gams %>%
ungroup() %>%
filter(variable=="GAI") %>%
select(Motion.Exclusion.Level, p.value) %>%
group_by(Motion.Exclusion.Level)
hist_gai_p=ggplot(pdat, aes(x=p.value, fill=Motion.Exclusion.Level, color=Motion.Exclusion.Level))+
geom_histogram(position = "identity", alpha=0.5, inherit.aes=TRUE, binwidth = .05)+
scale_x_continuous(expand = c(0, 0))+
scale_y_continuous(expand = c(0, 0))+
labs(x='p value', y='')+
scale_color_manual(labels=c('Strict', 'Lenient'), values = c("#f55154", "#9FB0CC"))+
scale_fill_manual(labels=c('Strict', 'Lenient'), values = c("#f55154", "#9FB0CC"))+
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), panel.background = element_blank(),
axis.line = element_line(size=.5), panel.border = element_blank())+
My_Theme+
theme(axis.title.y = element_blank())+
ggtitle("GAI")+
theme(plot.title = element_text(size = 9, hjust = 0.5))+
theme(legend.position = "none")+
theme(axis.title.x = element_text(size = 7))
hist_p_legend <- cowplot::get_legend(hist_gai_p + guides(color = guide_legend(nrow = 1))+theme(legend.position = "bottom", legend.text = element_text(size = 11), legend.key.size=unit(.15, "in"), legend.title = element_blank()))Combine histograms and print
fc_hist <- cowplot::plot_grid(hist_ados_p, hist_srs_p, hist_in_p, hist_hi_p, hist_mo_p, hist_age_p, hist_gai_p, ncol=7, rel_widths=c(1.18/7, .97/7, .97/7, .97/7, .97/7, .97/7, .97/7))
png("./CovariatesAndRS-fMRIUsability/fig_hist_rfc_cc.png",width=8,height=3,units="in",res=200)
cowplot::plot_grid(fc_hist, hist_p_legend, nrow=2, rel_heights=c(1, .1))
dev.off()## quartz_off_screen
## 2cowplot::plot_grid(fc_hist, hist_p_legend, nrow=2, rel_heights=c(1, .1))
Figure 6. Some covariates related to rs-fMRI exclusion probability are also related to functional connectivity. *Histograms of p values for generalized additive models of the relationship between edgewise functional connectivity in participants with usable rs-fMRI data and (from left to right) ADOS, social responsiveness scale (SRS) scores, inattentive symptoms, hyperactive/impulsive symptoms, total motor overflow as assessed during the Physical and Neurological Exam for Subtle Signs, age, and general ability index (GAI). For a given covariate, a clustering of p values near zero suggests that covariate is associated with functional connectivity for a greater number of edges. Several covariates appear to be related to functional connectivity using both the lenient motion quality control (slate blue bins) and the strict motion quality control (red bins).
Create table summarizing number of edges showing a nominally significant relationship (p<0.05 uncorrected) with each covariate, stratified by Primary Diagnosis
fc_gams_pCounts <- fc_gams %>%
select(variable, Motion.Exclusion.Level, edge, p.value) %>%
ungroup() %>%
group_by(variable, Motion.Exclusion.Level) %>%
summarise(pCount = sum(p.value<.05))## `summarise()` has grouped output by 'variable'. You can override using the
## `.groups` argument.fc_gams_pCounts$clustering = rep("no clustering", nrow(fc_gams_pCounts))
fc_gams_pCounts$clustering[fc_gams_pCounts$pCount>=10]="some clustering"
fc_gams_pCounts$clustering[fc_gams_pCounts$pCount>20]="clustering"
fc_gams_pCounts## # A tibble: 14 × 4
## # Groups: variable [7]
## variable Motion.Exclusion.Level pCount clustering
## <fct> <fct> <int> <chr>
## 1 ADOS Strict 21 clustering
## 2 ADOS Lenient 22 clustering
## 3 SRS Strict 12 some clustering
## 4 SRS Lenient 12 some clustering
## 5 Motor Overflow Strict 12 some clustering
## 6 Motor Overflow Lenient 10 some clustering
## 7 Inattention Strict 8 no clustering
## 8 Inattention Lenient 19 some clustering
## 9 Hyperactivity Strict 11 some clustering
## 10 Hyperactivity Lenient 17 some clustering
## 11 Age Strict 8 no clustering
## 12 Age Lenient 16 some clustering
## 13 GAI Strict 12 some clustering
## 14 GAI Lenient 9 no clustering#xtable(fc_gams_pCounts[order(fc_gams_pCounts$Motion.Exclusion.Level),])NOTE: I used the following arbitrary cutoffs to describe Figure 6 in the section 3.1.4
no clustering < 10 edges
some evidence of clustering: between 10-20 edges
clustering: >20 edges
Extra. Impact of motion QC on framewise displacement metrics
#mean and max FD are different for lenient and strict because some frames at the beginning and/or end of the scan are excluded for scans to pass lenient motion QC
dat3$MeanFD.None = dat3$MeanFramewiseDisplacement
dat3$MaxFD.None = dat3$MaxFramewiseDisplacement
#same for all levels
dat3$FramesWithFDLessThanOrEqualTo250microns.None = dat3$FramesWithFDLessThanOrEqualTo250microns
dat3$FramesWithFDLessThanOrEqualTo250microns.KKI = dat3$FramesWithFDLessThanOrEqualTo250microns
meanFD = c("ID", "MeanFramewiseDisplacement", "MeanFramewiseDisplacement.KKI",
"MeanFD.None")
maxFD = c("ID", "MaxFramewiseDisplacement", "MaxFramewiseDisplacement.KKI", "MaxFD.None")
framesFD = c("ID", "FramesWithFDLessThanOrEqualTo250microns",
"FramesWithFDLessThanOrEqualTo250microns.KKI",
"FramesWithFDLessThanOrEqualTo250microns.None")
fdID = c("ID")
meanFD.df <- reshape2::melt(dat3[, meanFD],
id.vars=names(dat3)[which(names(dat3) %in% fdID)],
variable.name = "Motion.Exclusion.Level",
value.name = "MeanFramewiseDisplacement")
levels(meanFD.df$Motion.Exclusion.Level)## [1] "MeanFramewiseDisplacement" "MeanFramewiseDisplacement.KKI"
## [3] "MeanFD.None"#rename levels to match motion QC levels in completeCases
levels(meanFD.df$Motion.Exclusion.Level) <- c("Strict", "Lenient", "None")
#repeat for MaxFD
maxFD.df <- reshape2::melt(dat3[, maxFD],
id.vars=names(dat3)[which(names(dat3) %in% fdID)],
variable.name = "Motion.Exclusion.Level",
value.name = "MaxFramewiseDisplacement")
levels(maxFD.df$Motion.Exclusion.Level)## [1] "MaxFramewiseDisplacement" "MaxFramewiseDisplacement.KKI"
## [3] "MaxFD.None"#rename levels to match motion QC levels in completeCases
levels(maxFD.df$Motion.Exclusion.Level) <- c("Strict", "Lenient", "None")
#merge meanFD.df and maxFD.df
fdMerg <- merge(meanFD.df, maxFD.df)
#repeat for FramesWithFDLessThanOrEqualTo250microns
frames.df <- reshape2::melt(dat3[, framesFD],
id.vars=names(dat3)[which(names(dat3) %in% fdID)],
variable.name = "Motion.Exclusion.Level",
value.name = "FramesWithFDLessThanOrEqualTo250microns")
levels(frames.df$Motion.Exclusion.Level)## [1] "FramesWithFDLessThanOrEqualTo250microns"
## [2] "FramesWithFDLessThanOrEqualTo250microns.KKI"
## [3] "FramesWithFDLessThanOrEqualTo250microns.None"#rename levels to match motion QC levels in qcMelt
levels(frames.df$Motion.Exclusion.Level) <- c("Strict", "Lenient", "None")
#merge with fdMerg
fdMerg <- merge(fdMerg, frames.df)
#merge with completeCases
completeCases <- merge(completeCases, fdMerg)
passOnly <- filter(completeCases, Included=="Included")Filter out “None” from Motion.Exclusion.Level to make remaining group differences in FD metrics following motion QC easier to see
mean framewise displacement
passOnly <- filter(passOnly, Motion.Exclusion.Level!="None")
meanFD_violin <- ggplot(passOnly, aes(Motion.Exclusion.Level, MeanFramewiseDisplacement,
fill = PrimaryDiagnosis, color=PrimaryDiagnosis)) +
geom_split_violin(trim=TRUE, alpha = 0.5, inherit.aes = TRUE, adjust = 2) +
stat_summary(fun = "mean", position = position_dodge(width = 0.18),
color="black", geom="point", aes(y=MeanFramewiseDisplacement))+
stat_summary(fun.data = "mean_cl_boot", position = position_dodge(width = 0.18),
color="black", geom="errorbar", width=.15)+
scale_fill_manual(values = c("#009E73", "#FDE599"))+
scale_color_manual(values = c("#05634a", "#E9D38D"))+
My_Theme_prop+theme(legend.title =element_blank())+
theme(axis.text.y=element_text(size=8))+
theme(legend.text = element_text(size = 7))+
theme(axis.title.y = element_blank())+
theme(axis.title.x = element_blank())+
theme(legend.position = c(0.35, .7))+
theme(legend.key.size=unit(.15, "in"))+
theme(legend.box.margin = margin(-2, -2, -2, -2))+
ggtitle("Mean FD")+
theme(plot.title = element_text(size = 8, hjust=0.5))max framewise displacement
maxFD_violin <- ggplot(passOnly, aes(Motion.Exclusion.Level, MaxFramewiseDisplacement,
fill = PrimaryDiagnosis, color=PrimaryDiagnosis)) +
geom_split_violin(trim=TRUE, alpha = 0.5, inherit.aes = TRUE, adjust = 2) +
stat_summary(fun = "mean", position = position_dodge(width = 0.18), color="black",
geom="point", aes(y=MaxFramewiseDisplacement))+
stat_summary(fun.data = "mean_cl_boot", position = position_dodge(width = 0.18),
color="black", geom="errorbar", width=.15)+
scale_fill_manual(values = c("#009E73", "#FDE599"))+
scale_color_manual(values = c("#05634a", "#E9D38D"))+
My_Theme_prop+
theme(axis.text.y=element_text(size=8))+
theme(axis.title.y = element_blank())+
theme(axis.title.x = element_blank())+
theme(legend.position = "none")+
ggtitle("Max FD")+
theme(plot.title = element_text(size = 8, hjust=0.5))frames with FD <= 0.25 mm
frames_violin <- ggplot(passOnly, aes(Motion.Exclusion.Level, FramesWithFDLessThanOrEqualTo250microns,
fill = PrimaryDiagnosis, color=PrimaryDiagnosis)) +
geom_split_violin(trim=TRUE, alpha = 0.5, inherit.aes = TRUE, adjust = 2) +
stat_summary(fun = "mean", position = position_dodge(width = 0.18), color="black",
geom="point", aes(y=FramesWithFDLessThanOrEqualTo250microns))+
stat_summary(fun.data = "mean_cl_boot", position = position_dodge(width = 0.18),
color="black", geom="errorbar", width=.15)+
scale_fill_manual(values = c("#009E73", "#FDE599"))+
scale_color_manual(values = c("#05634a", "#E9D38D"))+
My_Theme_prop+
theme(axis.text.y=element_text(size=8))+
theme(axis.title.y = element_blank())+
theme(axis.title.x = element_blank())+
theme(legend.position = "none")+
ggtitle("Frames with FD<0.25 mm")+
theme(plot.title = element_text(size = 8, hjust=0.5))Combine 3 FD plots with just lenient and strict Motion.Exclusion.Levels and save
## quartz_off_screen
## 2
Framewise displacement metrics are more similar across diagnosis groups using the strict motion QC, but very few participants are labeled as usable.